Dyno Lab — an AI safety, alignment and interpretability research workbench for Apple Silicon. Explore activations, probes, SAEs and interventions with local inference, a Python SDK, APIs and MCP.
- ai
- ai-interpretability
- ai-safety
- aisafety
132,241 visible repositories in the index
697–720 of 753 · Most stars
Dyno Lab — an AI safety, alignment and interpretability research workbench for Apple Silicon. Explore activations, probes, SAEs and interventions with local inference, a Python SDK, APIs and MCP.
llama.cpp fork for significantly improved performance on Ampere generation Nvidia cards (especially RTX 3090 / 3090 Ti): example: 110+ tok/s over a 100K-token generation at temperature 1 for qwen3.8, supporting full 262K ctx. Supports low memory (12gb) configs for 200k+ ctx. 120+ TPS, 3300+ prefill for QFN 3.8.
TranslatedThe operational superset of Pi Coding Agent — everything Pi, plus observability, governance, recovery, evaluation and multi-agent orchestration.
TranslatedBig Blue Fish desktop pet with local LLM. A Windows AI desktop pet with a built-in 4B local large model. It uses a whale-girl persona that combines weather, tech news and the current application to keep you gentle company through text bubbles. Supports one-click installation and pausing inference when off power.
A team of Hermes agents you chat with, each with its own screen on one Orgo computer. Composio apps, xAI voice and X research.
Language model for DOS that fits on a floppy disk
🤖 Bring the power of ChatGPT to Microsoft Word
TranslatedLocal-first prompt vault for Windows · Tauri 2 + React 19 + TypeScript · Local-first desktop tool for collecting prompts
OpenAI (and DeepSeek, Azure OpenAI, YandexGPT, Ollama, GigaChat, Qwen) API wrapper for Delphi. Use ChatGPT, DALL-E, Whisper and other products.
A customizable Windows desktop island for music, focus, live lyrics and your own AI companion.
Berth is a lightweight GPU resource manager written in Go. It controls which applications get access to your GPU, when they get it, and how that access is handed off.
Pure-Rust inference engine for Qwen GGUF models (Qwen3 to Qwen3.8, dense and MoE up to 770B+). CUDA and Vulkan (NVIDIA, AMD), hybrid CPU+GPU experts, MTP speculative decoding, OpenAI and Anthropic compatible server. Up to ~1.7x llama.cpp decode on a 12 GB RTX 3060.
A new desktop/web app experience for buzz
Kubernetes operator for self-hosted LLM inference across a heterogeneous GPU fleet: NVIDIA CUDA, AMD Vulkan, and Apple Silicon Metal. Runtimes: llama.cpp, vLLM, TGI, mlx-server. Multi-GPU sharding, model caching, OpenAI-compatible endpoints. Apache-2.0, run across homelab and on-prem fleets, actively developed.
Fast local LLM inference for the GPU you already have: one Rust binary, GGUF models, an OpenAI-compatible server and a chat page with a ghost.
TranslatedAcademic research library for DSH — search, cite, approve imports, hand off notes. Also: a research literature library plugin for DeepSeek Harness: search papers, include sources, approve imports into the library, and hand off note leads to other plugins in the same session.
Go AI SDK, the Go way. One unified API across 21+ providers. Streaming, structured output, MCP support, stdlib only. Go AI SDK for AI applications inspired by Vercel AI SDK.
llama.cpp fork that abliterates at inference
Cross-platform desktop LLM client for OpenAI ChatGPT, DeepSeek R1, Anthropic Claude and more, with a focus on privacy and security.
Self-hosted, multi-tenant agent harness for SRE incident diagnosis — risk-gated tool approvals, durable resumable runs, audit trail, model gateway, MCP tools, and a native desktop chat client. Written in Go.
Structured outputs, vectorization and tool calling for your Ash application
Enterprise-grade benchmark evaluating whether AI agents deliver verifiable business outcomes in real business systems — general domain, finance, supply chain, and manufacturing.
M-Courtyard: Local AI Model Fine-tuning Assistant for Apple Silicon. Zero-code, zero-cloud, privacy-first desktop app powered by Tauri + React + mlx-lm.
One command downloads a complete offline, no-internet emergency library: local AI (CPU/GPU), document Q&A, full Wikipedia, offline maps for any country/region (default US), survival/repair/medical archives, and public-domain manuals. Asks only where to save.